Imputacija visokodimenzionalnih bioloških podatkov

Jelovčan, Gašper (author), Curk, Tomaž (mentor)

V diplomskem delu smo predstavili metode za imputacijo visokodimenzionalnih bioloških podatkov pridobljenih z metodami sekvenciranja posameznih celic. Pridobljenim podatkom mnogokrat manjkajo vrednosti, ki jih poskušamo nadomestiti z ocenami vrednosti. Preizkusili smo različne metode za imputacijo. Implementirali smo jih v modulu v programskem jeziku Python. Metode smo preverili na umetnih in pravih podatkih. Na preizkušenih podatkih so vse metode dosegle dobre rezultate. Najbolj ustrezna je bila metoda pCMF.

• Language: Slovenian
• Keywords: Posamezne celice, števni podatki, redki podatki, imputacija.
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2019
• PID: 20.500.12556/RUL-109866
• COBISS.SI-ID: 1538319811
• Publication date in RUL: 09.09.2019

Secondary language

• Language: English
• Title: Imputation of high-dimensional biological data

Abstract

In this thesis we present methods for solving problems of high-dimensional biological data imputation collected by sequencing individual cells. We try to assign values to the missing data, replacing them with estimations. We tried several imputation methods. We have implemented imputation methods as a module in programming language Python. Then we tested them using synthetic data and real biological data. The evaluation showed that all methods achieve good results. The pCMF method performed the best.

• Keywords: Single cell, count data, sparse data, imputation.